# Also in the Article

Ensemble generation
This protocol is extracted from research article:
Dynamic tuning of FRET in a green fluorescent protein biosensor

Procedure

First, we individually modeled all possible backbone dihedral angles (ϕ and ψ; sampling in steps of 60°) of the linker residues (Arg229 to Gln231 for mCerulean3 and Met311 to Gly313 for cpVenus), which did not result in a steric clash between one of the modified fluorescent protein domains and the TnC domain. Each of the two linker regions was modeled independently. Of all possible ϕ, ψ combinations (117,649), the rotation of mCerulean3 resulted in 477 possible conformations, while cpVenus allowed 84 conformations, providing altogether 40,086 possible conformations.

Second, we combined the possible structures for mCerulean-TnC and TnC-cpVenus at random in ensembles of six members. We sampled randomly 1 million six-membered ensembles out of the 40,068 possible ϕ, ψ combinations of the linkers between mCerulean and TnC and between TnC and cpVenus to ensure a proper exploration of the conformational space of Twitch.

Third, we calculated the RDC range (using the tensor derived as described above) and FRET of the ensembles and compared them with the experimental values by defining an ensemble quality factor$Qens=∑i=1i=6(RDCi−RDCeRDCe)2+(FRETi−FRETeFRETe)2$where RDCi are distribution ranges, the subindex e indicates an experimental value, and i indicates the value from an ensemble member. Such quality factor value is different from 0 if the agreements of the predicted RDC and FRET responses from the ensemble deviate from the experimental data and it is 0 in the case of complete agreement. Last, the ensembles were sorted by their Qens, and the lowest one was selected.

Note: The content above has been extracted from a research article, so it may not display correctly.

Q&A